Claude Code
Add the server:
claude mcp add --transport http minefort https://minefort.com/mcp
Then run this inside Claude Code and sign in to minefort:
/mcp
Connect Claude Code, Codex or opencode to Minefort. Ask it to start a server, fix a config or build an experience, and it works on the real thing.
Server URL
The first time, your tool opens a browser to sign in to Minefort and approve access. It can only reach what your account can manage.
Add the server:
claude mcp add --transport http minefort https://minefort.com/mcp
Then run this inside Claude Code and sign in to minefort:
/mcp
Add the server:
codex mcp add minefort --url https://minefort.com/mcp
Then sign in:
codex mcp login minefort
Open the link and approve it in Cursor. The first time, it asks you to sign in to Minefort.
Open the link and approve it in VS Code. The first time, it asks you to sign in to Minefort.
Add this to opencode.json:
{
"mcp": {
"minefort": {
"type": "remote",
"url": "https://minefort.com/mcp"
}
}
}Then sign in:
opencode mcp auth minefort
Cursor, VS Code, a Claude Desktop custom connector, or any client with remote MCP and OAuth: add the server URL.
https://minefort.com/mcp
Your agent picks these tools on its own. Ask for what you want in plain words.
List, start and stop the servers and experiences you manage.
Read and edit files, tail the console and run commands.
Upload local files and assets through short-lived curl URLs.
Build Citadel experience packs. Edits to .luau scripts reload in about a second, and the console shows reloaded or the script error.
Check a build for faults, render it from every side and look at its models, without joining the game.
Put an experience live when it is ready.
Put items and passes on sale for Gems, and earn Creator Rewards on every sale.
Take a backup before a risky change, and restore it if something goes wrong.
When the engine or a tool falls short, your agent can tell Minefort. That is how both grow.